New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

enhancement: 优化rearrange算子 #8

Draft

pwhMass wants to merge 1 commit into YdrMaster:main from pwhMass:rearrange

pwhMass commented Jan 20, 2025

No description provided.

Author

pwhMass commented Jan 20, 2025

目前代码不可用，且只针对2维做了转置优化，正在测试什么情况下性能更好

YdrMaster self-assigned this

YdrMaster marked this pull request as draft

January 20, 2025 09:37

pwhMass force-pushed the rearrange branch 2 times, most recently from cfd19d9 to b6d2b42 Compare

January 24, 2025 13:43

pwhMass force-pushed the rearrange branch from 620a20b to bf578b3 Compare

February 9, 2025 05:54

pwhMass force-pushed the rearrange branch from 1be9bd6 to de841a0 Compare

February 20, 2025 16:16


          ehancement(cuda): 优化rearrange算子，并使其支持多维张量

8d4cd02

需要注意目前 ARRAY_SIZE 的大小是5，该常亮与可接受的Tensor的维度有关，但太大会导致kernel计算量增大
Operator 需要用到max_warps_block，warp_size来辅助计算，目前并未用到
block_size 目前固定位256，可进一步优化

pwhMass force-pushed the rearrange branch from de841a0 to 8d4cd02 Compare

February 20, 2025 16:23

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet